Indus script

Early examples of the symbol system are found during the Early Harappan period of the Indus civilisation, dated to possibly as early as the 35th century BCE.[22][23] Pottery inscriptions and clay impressions of inscribed Harappan seals have also been found from the Kot Diji phase of the Early Harappan period, dated to c. 2800–2600 BCE.[24]

In the Mature Harappan period, from about c. 2600–1900 BCE, strings of Indus signs are commonly found on flat, rectangular stamp seals as well as written or inscribed on a multitude of other objects including pottery, tools, tablets, and ornaments. Signs were written using a variety of methods including carving, chiselling, embossing, and painting applied to diverse materials such as terracotta, sandstone, soapstone, bone, shell, copper, silver, and gold.[25] As of 1977, Iravatham Mahadevan noted, that about 90% of the Indus script seals and inscribed objects discovered so far were found at sites in Pakistan along the Indus river and its tributaries, such as Mohenjo-daro and Harappa,[lower-alpha 1] while other sites located elsewhere account for the remaining 10%.[lower-alpha 2][26][27] Often, animals such as bulls, water buffaloes, elephants, rhinoceros, and the mythical "unicorn"[lower-alpha 3] accompanied the text on seals to help the illiterate identify the origin of a particular seal.[29]

The Late Harappan period followed the more urbanised Mature Harappan period, and was a period of fragmentation and localisation which preceded the early Iron Age in the Indian subcontinent. Inscriptions have been found at sites associated with the localised phases of this period. The Jhukar phase of the Late Harappan period was centred on the present-day province of Sindh, Pakistan. Although seals from this phase lack the Indus script which characterised the preceding phase of the civilisation, some potsherd inscriptions have been noted.[30] Both seals and potsherds bearing Indus script text, dated to c. 2200–1600 BCE, have been found at sites associated with the Daimabad culture of the Late Harappan period, in present-day Maharashtra.[31]

Numerous artefacts, particularly potsherds and tools, bearing markings inscribed into them have been found in Central India, South India, and Sri Lanka dating to the Megalithic Iron Age which followed the Late Harappan period. These markings include inscriptions in the Brahmi and Tamil-Brahmi scripts, but also include non-Brahmi graffiti symbols which co-existed contemporaneously with the Tamil-Brahmi script.[32] As with the Indus script, there is no scholarly consensus on the meaning of these non-Brahmi symbols. Some scholars, such as the anthropologist Gregory Possehl,[4] have argued that the non-Brahmi graffiti symbols are a survival and development of the Indus script into and during the 1st millennium BCE.[32] In 1960,[33] archaeologist B. B. Lal found that a majority[lower-alpha 4] of the megalithic symbols he had surveyed were identifiably shared with the Indus script, concluding that that there was a commonness of culture between the Indus Valley Civilisation and the later Megalithic period.[34] Similarly, Indian epigraphist Iravatham Mahadevan has argued that sequences of Megalithic graffiti symbols have been found in the same order as those on comparable Harappan inscriptions and that this is evidence that language used by the Iron Age people of south India was related to or identical with that of the late Harappans.[35][15][36]

Variations of 'sign 4';[lower-alpha 5] such variation makes distinguishing signs from allographical variants difficult, and scholars have proposed different ways to classify elements of the Indus script.[38]

Most scholars agree that the Indus script was generally written from right to left,[51][41][52] however, some exceptions wherein the script is written left to right or in a boustrophedon mode are also known.[51] Although the script is undeciphered, the writing direction has been deduced from external evidence, such as instances of the symbols being compressed on the left side as if the writer is running out of space at the end of the row.[51][53] As almost all these signs are used on seals, which create a mirror image impression on the clay or ceramic on which the seal is affixed, the writing on the seal has to be reversed (as is AMBULANCE on the front of an emergency vehicle, to be read in a rear-view mirror). Hence, it cannot be deduced that the language itself was written from right to left.

Scholars have compared the Indus valley script to the Brahmi and Tamil-Brahmi scripts, suggesting that there may be similarities between them. These similarities were first suggested by early European scholars, such as the archaeologist John Marshall[54] and the Assyriologist Stephen Langdon,[55] with some, such as G.R. Hunter,[9] proposing an indigenous origin of Brahmi with a derivation from the Indus script.

Scholars have also compared the Indus valley script with the Linear Elamite writing system used in Elam, an ancient Pre-Iranian civilisation that was contemporaneous with the Indus Valley civilisation. The two writing systems were contemporary to each other. On comparing the Elamite script to the Indus script, a number of similar symbols have been found.[56]

The following factors are usually regarded as the biggest obstacles to successful decipherment:

• The average length of the inscriptions is around five signs,[58] and the longest only 34 characters long, found on a copper plate belonging to the mature Harappan period.[57]
• There are doubts whether the Indus script records a written language or is instead a system of non-linguistic signs or proto-writing similar to merchant's marks and house marks, and to the contemporary accounting tokens and numerical clay tablets of Mesopotamia.[43]
• Assuming it is a written language, the underlying Harappan language has not been identified,[5] though some 300 loanwords in the Rigveda are a good starting point for comparison.[lower-alpha 9][60]
• No bigraphic or bilingual texts, like a Rosetta Stone, have been found.[5]
• No names of Indus rulers or personages are known to be attested in surviving historical records or myths, as was the case with rulers like Rameses and Ptolemy, who were known to hieroglyphic decipherers from records attested in Greek.[5][lower-alpha 10]

Over the years, numerous decipherments have been proposed, but there is no established scholarly consensus.[43][61] The few points on which there exists scholarly consensus are the right-to-left direction of the majority of the inscriptions,[41][5] numerical nature of certain stroke-like signs,[41][5] functional homogeneity of certain terminal signs,[41] and some generally adopted techniques of segmenting the inscriptions into initial, medial, and terminal clusters.[41] Over 100 (mutually exclusive) attempts at decipherment have been published since the 1920s,[62][5] and the topic is popular among amateur researchers.[lower-alpha 11]

The Indus script 'fish sign', associated with the Dravidian reading mīn, has been interpreted as its homophone, meaning "star", per the rebus principle in the context of some Indus inscriptions[63]

Although no clear consensus has been established, there are those who argue that the Indus script recorded an early form of a Dravidian language.[43] Early proponents included the archaeologist Henry Heras, who suggested several readings of signs based on a proto-Dravidian assumption.[64]

Based on computer analysis,[65] the Russian scholar Yuri Knorozov suggested that a Dravidian language is the most likely candidate for the underlying language of the script.[66] The Finnish scholar Asko Parpola led a Finnish team in the 1960s–80s that, like Knorozov's Soviet team, worked towards investigating the inscriptions using computer analysis. Parpola similarly concluded that the Indus script and Harappan language "most likely belonged to the Dravidian family".[67] A comprehensive description of Parpola's work up to 1994 is given in his book Deciphering the Indus Script.[65] Supporting this work, the archaeologist Walter Fairservis argued that Indus script text on seals could be read as names, titles, or occupations, and suggested that the animals depicted were totems indicating kinship or possibly clans.[43][68][69] The computational linguist Rajesh P. N. Rao, along with a team of colleagues, performed an independent computational analysis and concluded that the Indus script has the structure of a written language, supporting prior evidence for syntactic structure in the Indus script, and noting that the Indus script appears to have a similar conditional entropy to Old Tamil.[70][71]

These scholars have proposed readings of many signs; one such reading was legitimised when the Dravidian homophonous words for 'fish' and 'star', mīn, were hinted at through drawings of both the things together on Harappan seals.[72] In a 2011 speech, Rajesh P. N. Rao said that Iravatham Mahadevan and Asko Parpola "have been making some headway on this particular problem", namely deciphering the Indus script, but concluded that their proposed readings, although they make sense, are not yet proof.[73]

Indus script on a stamp seal depicting a buffalo-horned figure surrounded by animals, dubbed the 'Lord of the Beasts' or 'Paśupati' seal (c. 2350–2000 BCE).[lower-alpha 12]

In his 2014 publication Dravidian Proof of the Indus Script via The Rig Veda: A Case Study, the epigraphist Iravatham Mahadevan identified a recurring sequence of four signs which he interpreted as an early Dravidian phrase translated as "Merchant of the City".[75] Commenting on his 2014 publication, he stressed that he had not fully deciphered the Indus script, although he felt his effort had "attained the level of proof" with regards to demonstrating that the Indus script was a Dravidian written language.[76]

Perhaps the most influential proponent of the hypothesis that the Indus script records an early Indo-Aryan language is the Indian archaeologist Shikaripura Ranganatha Rao,[43] who in his books, Lothal and the Indus Civilization and The Decipherment of the Indus Script, wrote that he had deciphered the script. While dismissing most such attempts at decipherment, John E. Mitchiner commented that "a more soundly-based but still greatly subjective and unconvincing attempt to discern an Indo-European basis in the script has been that of Rao".[77] S.R. Rao perceived a number of similarities in shape and form between the late Harappan characters and the Phoenician letters, and argued that the Phoenician script evolved from the Harappan script, and not, as the classical theory suggests from the Proto-Sinaitic script.[43][78] He compared it to the Phoenician alphabet, and assigned sound values based on this comparison.[43] Reading the script from left to right, as is the case with Brahmi, he concluded that Indus inscriptions were "Sanskritic", including the numerals aeka, dwi, tra, chatus, panta, happta/sapta, dasa, dvadasa, shata (1, 2, 3, 4, 5, 7, 10, 12, 100).[79]

S.R. Rao's interpretation helped to bolster Hindu nationalist and Aryan indigenist views propagated by writers, such as David Frawley, who hold the conviction that Indo-Aryan peoples are the original Bronze Age inhabitants of the Indian subcontinent and that the Indo-European language family originated in India.[43] However, there are many problems with this hypothesis, particularly the cultural differences evident between the Indus River Civilisation and Indo-European cultures, such as the role of horses in the latter; as Parpola put it, "there is no escape from the fact that the horse played a central role in the Vedic and Iranian cultures".[80]

A less popular hypothesis suggests that the Indus script belongs to the Munda family of languages. This language family is spoken largely in Eastern India, and is related to some Southeast Asian languages. However, much like the Indo-Aryan language, the reconstructed vocabulary of early Munda does not reflect the Harappan culture,[81] therefore, its candidacy for being the language of the Indus Civilisation is dim.[82]

Indus script tablet recovered from Khirasara, Indus Valley

A sequence of Indus characters from the northern gate of Dholavira, dubbed the Dholavira Signboard

An opposing hypothesis is that these symbols are nonlinguistic signs which symbolise families, clans, gods, and religious concepts, and are similar to components of coats of arms or totem poles. In a 2004 article, Steve Farmer, Richard Sproat, and Michael Witzel presented a number of arguments stating that the Indus script is nonlinguistic.[83] The main ones are the extreme brevity of the inscriptions, the existence of too many rare signs (which increase over the 700-year period of the Mature Harappan civilisation), and the lack of the random-looking sign repetition that is typical of language.[84]

Asko Parpola, reviewing the Farmer et al. thesis in 2005, stated that their arguments "can be easily controverted".[85] He cited the presence of a large number of rare signs in Chinese and emphasised that there was "little reason for sign repetition in short seal texts written in an early logo-syllabic script". Revisiting the question in a 2008 lecture,[86] Parpola took on each of the 10 main arguments of Farmer et al., presenting counterarguments for each.

A 2009 paper[70] published by Rajesh P.N. Rao, Iravatham Mahadevan and others in the journal Science also challenged the argument that the Indus script might have been a nonlinguistic symbol system. The paper concluded that the conditional entropy of Indus inscriptions closely matched those of linguistic systems like the Sumerian logo-syllabic system, Rig Vedic Sanskrit etc., but they are careful to stress that by itself does not imply that the script is linguistic. A follow-up study presented further evidence in terms of entropies of longer sequences of symbols beyond pairs.[87] However, Sproat argued that there existed a number of misunderstandings in Rao et al., including a lack of discriminative power in their model, and argued that applying their model to known non-linguistic systems such as Mesopotamian deity symbols produced similar results to the Indus script. Rao et al.'s argument against Sproat's arguments and Sproat's reply were published in Computational Linguistics in December 2010.[88][71] The June 2014 issue of Language carries a paper by Sproat that provides further evidence that the methodology of Rao et al. is flawed.[89] Rao et al.'s rebuttal of Sproat's 2014 article and Sproat's response are published in the December 2015 issue of Language.[90][91]